Shulou(Shulou.com)06/01 Report--

This article mainly introduces the manual compilation and installation of ceph jewe optimization skills, has a certain reference value, interested friends can refer to, I hope you can learn a lot after reading this article, the following let the editor with you to understand.

First, install ceph manually.

Manually install the jewel version of ceph to the hardware server according to http://my.oschina.net/linuxhunter/blog/682013.

Second, the method of testing ceph cluster.

Use the rados bench command that comes with ceph to simply test the performance of manually building a ceph cluster, and the tool to view system performance uses the perf command. Since the default installed ubuntu environment does not have perf tools and their dependencies installed, you need to install perf tools manually. # apt-get install perf linux-tools-4.4.0-21-generic. After installation, enable the two terminals to run the # perf top command and the # rados banch-p test_rbd 60 write-- no-cleanup command, respectively.

Third, the problems found.

On the terminal running the # perf top command, it is found that ceph-osd takes up 35% of the CPU time in calling the ceph_crc32_sctp function, and the pressure on the cluster is not great at this time, so it is decided to find out why ceph_crc32_sctp takes up so much CPU time from the source code.

The ceph_crc32_sctp function is located in the src/common/sctp_crc32.c file, and only the ceph_choose_crc32 function calls the ceph_crc32_sctp function. It is not difficult to analyze the function ceph_choose_crc32, which chooses the method of calculating crc32 based on the architecture of the current CPU. If CPU is the CPU of intel and supports the sse4 instruction set, use the hardware to calculate the crc32 value (call the ceph_crc32_intel_fast function), otherwise use the software to calculate the crc32 value (call the ceph_crc32_sctp function). The CPU we use is intel Xeon 2560 v2, which of course supports the sse4 instruction set.

Fourth, the way to solve the problem.

If you want ceph to support hardware calculation of crc32 values, install the yasm toolkit when compiling ceph so that ceph can access the hardware correctly through assembly statements. So you need to install the yasm toolkit manually, that is, # apt-get install yasm. Then recompile and install ceph manually according to http://my.oschina.net/linuxhunter/blog/682013.

Fifth, the optimization results.

Based on the ceph compiled by the yasm toolkit, after rados bench testing, it is found that the crc32_iscsi_00 function takes up about 8% of the CPU time in the ceph-osd process in perf top, and the crc32_iscsi_00 function is called by the ceph_crc32_intel_fast function. Therefore, the compiled ceph based on yasm will make full use of the hardware CPU feature to calculate the value of crc32, so the cpu occupancy will be much lower, from 35% to 8%.

Thank you for reading this article carefully. I hope the article "what are the optimization skills for manual compilation and installation of ceph jewe" shared by the editor will be helpful to you. At the same time, I also hope you will support us and pay attention to the industry information channel. More related knowledge is waiting for you to learn!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.